JH
Jonas Haslbeck
Author with expertise in Network Analysis of Psychopathology and Mental Disorders
Achievements
Cited Author
Open Access Advocate
Key Stats
Upvotes received:
0
Publications:
13
(77% Open Access)
Cited by:
1,163
h-index:
26
/
i10-index:
48
Reputation
Biology
< 1%
Chemistry
< 1%
Economics
< 1%
Show more
How is this calculated?
Publications
0

Using network analysis to examine links between individual depressive symptoms, inflammatory markers, and covariates

Eiko Fried et al.Oct 16, 2019
Abstract Background Studies investigating the link between depressive symptoms and inflammation have yielded inconsistent results, which may be due to two factors. First, studies differed regarding the specific inflammatory markers studied and covariates accounted for. Second, specific depressive symptoms may be differentially related to inflammation. We address both challenges using network psychometrics. Methods We estimated seven regularized Mixed Graphical Models in the Netherlands Study of Depression and Anxiety (NESDA) data ( N = 2321) to explore shared variances among (1) depression severity, modeled via depression sum-score, nine DSM-5 symptoms, or 28 individual depressive symptoms; (2) inflammatory markers C-reactive protein (CRP), interleukin 6 (IL-6), and tumor necrosis factor α (TNF- α ); (3) before and after adjusting for sex, age, body mass index (BMI), exercise, smoking, alcohol, and chronic diseases. Results The depression sum-score was related to both IL-6 and CRP before, and only to IL-6 after covariate adjustment. When modeling the DSM-5 symptoms and CRP in a conceptual replication of Jokela et al ., CRP was associated with ‘sleep problems’, ‘energy level’, and ‘weight/appetite changes’; only the first two links survived covariate adjustment. In a conservative model with all 38 variables, symptoms and markers were unrelated. Following recent psychometric work, we re-estimated the full model without regularization: the depressive symptoms ‘insomnia’, ‘hypersomnia’, and ‘aches and pain’ showed unique positive relations to all inflammatory markers. Conclusions We found evidence for differential relations between markers, depressive symptoms, and covariates. Associations between symptoms and markers were attenuated after covariate adjustment; BMI and sex consistently showed strong relations with inflammatory markers.
0

Productive explanation: A framework for evaluating explanations in psychological science.

Noah Dongen et al.Jul 18, 2024
The explanation of psychological phenomena is a central aim of psychological science. However, the nature of explanation and the processes by which we evaluate whether a theory explains a phenomenon are often unclear. Consequently, it is often unknown whether a given psychological theory indeed explains a phenomenon. We address this shortcoming by proposing a productive account of explanation: a theory explains a phenomenon to some degree if and only if a formal model of the theory produces the statistical pattern representing the phenomenon. Using this account, we outline a workable methodology of explanation: (a) explicating a verbal theory into a formal model, (b) representing phenomena as statistical patterns in data, and (c) assessing whether the formal model produces these statistical patterns. In addition, we provide three major criteria for evaluating the goodness of an explanation (precision, robustness, and empirical relevance), and examine some cases of explanatory breakdowns. Finally, we situate our framework within existing theories of explanation from philosophy of science and discuss how our approach contributes to constructing and developing better psychological theories. (PsycInfo Database Record (c) 2024 APA, all rights reserved).
0
Citation1
0
Save
0

Sensitivity analysis of prior distributions in Bayesian graphical modeling: Guiding informed prior choices for conditional independence testing

Nikola Sekulovski et al.Jan 1, 2024
Bayesian analysis methods provide a significant advancement in network psychometrics, allowing researchers to use the edge inclusion Bayes factor for testing conditional independence between pairs of variables in the network. Using this methodology requires setting prior distributions on the network parameters and on the network’s structure. However, the impact of both prior distributions on the inclusion Bayes factor is underexplored. In this paper, we focus on a specific class of Markov Random Field models for ordinal and binary data. We first discuss the different choices of prior distributions for the network parameters and the network structure, and then perform an extensive simulation study to assess the sensitivity of the inclusion Bayes factor to these distributions. We pay particular attention to the effect of the scale of the prior on the inclusion Bayes factor. To improve the accessibility of the results, we also provide an interactive Shiny app. Finally, we present the R package simBgms, which provides researchers with a user-friendly tool to perform their own simulation studies for Bayesian Markov Random Field models. All of this should help researchers make more informed, evidence-based decisions when preparing to analyze empirical data using network psychometric models.
0

Stimulant medication and symptom interrelations in children, adolescents and adults with attention-deficit/hyperactivity disorder

Zarah Pal et al.Nov 11, 2024
Abstract Stimulant medication is effective in alleviating overall symptom severity of attention-deficit/hyperactivity disorder (ADHD), yet interindividual variability in treatment response and tolerability still exists. While network analysis has identified differences in ADHD symptom relations, the impact of stimulant medication remains unexplored. Increased understanding of this association could provide valuable insights for optimizing treatment approaches for individuals with ADHD. In this study, we compared and characterized ADHD symptom networks (including 18 ADHD symptoms) between stimulant-treated ( n = 348) and untreated ( n = 70) individuals with ADHD and non-ADHD controls (NACs; n = 444). Moreover, we compared symptom networks between subgroups defined by their stimulant treatment trajectory (early-and-intense use, late-and-moderate use). Stimulant-treated individuals with ADHD showed stronger associations between symptoms, compared with untreated individuals with ADHD and NACs. We found no differences in symptom networks between the stimulant treatment trajectory subgroups. Prospective longitudinal studies are needed to disentangle whether the identified differences stem from treatment or pre-existing factors.
Load More